A New Classification Framework to Evaluate the Entity Profiling on the Web

Abstract

Recently, we have witnessed entity profiling (EP) becoming increasingly one of the most important topics in information extraction, personalized applications, and web data analysis. EP aims to identify, extract, and represent a compact summary of valuable information about an entity based on the data related to it. To determine how EP systems have developed, during the last few years, this article reviews EP systems through a survey of the literature, from 2000 to 2015. To fulfill this aim, we introduce a comparison framework to compare and classify EP systems. Our comparison framework is composed of thirteen criteria that include: profiling source, the entity being modeled, the information that constitutes the profile, representation schema, profile construction technique, scale, scope/target domain, language, updating mechanism, enrichment technique, dynamicity, evaluation method, and application among others. Then, using the comparison framework, we discuss the recent development of the field and list some of the open problems and main trends that have emerged in EP to provide a proper guideline for researchers to develop or use robust profiling systems with suitable features according to their needs.